Anthropic's revenue has overtaken that of OpenAI, and it is sprinting for the largest IPO in history.

For OpenAI, its strong rival Anthropic is no longer a mere "follower".

Recently, Anthropic informed potential investors that its revenue in the second quarter increased by at least 14 times year-on-year. Relevant documents show that Anthropic's preliminary revenue in the most recent full quarter exceeded 11.5 billion US dollars, compared with 787 million US dollars in the same period of 2025 and 4.73 billion US dollars in the first quarter of this year, with a quarter-on-quarter increase of over 140%.

In addition, the documents also show that the company recorded an adjusted operating profit of 559 million US dollars for the first time in the second quarter, achieving profitability at the operational level. Calculated on this basis, Anthropic's latest annualized revenue run rate has exceeded 46 billion US dollars.

On August 18, people familiar with the matter said that Anthropic's annualized revenue calculated based on current business performance is expected to exceed 65 billion US dollars, surging more than six times from the level at the end of last year.

In contrast, OpenAI's revenue in the second quarter was 6.7 billion US dollars, a quarter-on-quarter increase of only 18% compared with 5.7 billion US dollars in the first quarter. The growth rate slowed down significantly, while operating losses continued to expand in the same period, and it failed to turn profitable. In 2026, its annualized revenue is expected to exceed 40 billion US dollars, doubling its 2025 revenue.

One side achieved quarterly profitability, while the other side suffered high losses. The two report cards with diverging trends indicate that the business models of the two leading AI companies have undergone substantial differentiation.

Institutional analysts believe that the widening revenue gap between the two is mainly due to changing choices brought by the actual usage costs of enterprise customers. Although the listed price of a single call of Anthropic's flagship model is not low, its higher task completion accuracy reduces repeated calls and manual reviews, giving enterprises an advantage in actual comprehensive usage costs.

01

All along, Anthropic has concentrated its resources on enterprise development scenarios. Its Claude Code programming tool has achieved explosive growth, firmly capturing the needs of developers and technology-oriented enterprises, and the enterprise-side business contributes the vast majority of incremental revenue. OpenAI has built a huge consumer-side user base relying on ChatGPT, but consumer subscription growth has peaked, and the expansion speed of its enterprise business lags behind its rival.

Moreover, Anthropic attaches greater importance to revenue quality and gross margin, prioritizing the development of enterprise tool products that can be monetized quickly. OpenAI maintains a strategy of full investment in large models and multiple product lines, with a huge scale of investment in computing power infrastructure. For every dollar of revenue generated, it has to bear high hardware and reasoning costs.

In April this year, according to reports from multiple media outlets, OpenAI failed to meet its monthly sales targets for several months, and ChatGPT failed to achieve the goal of 1 billion weekly active users by the end of 2025. Due to the rising popularity of Google Gemini last year, the churn rate of ChatGPT's subscribing users still poses a challenge.

Sarah Friar, OpenAI's Chief Financial Officer, also expressed concerns in conversations with other company executives that if OpenAI's sales growth is not fast enough, it may not be able to afford future computing power demands.

While the performance landscape is changing, both companies are stepping up preparations for IPOs, hoping to use funds from the public market to support future computing power and R&D investment. Anthropic has submitted a confidential S-1 prospectus document to the U.S. SEC in June this year, with Morgan Stanley, Goldman Sachs, and JPMorgan Chase acting as joint underwriters, targeting an IPO window between September and October this year.

Some investors have given Anthropic a maximum valuation expectation of 2 trillion US dollars. If this target is finally fulfilled, Anthropic will surpass SpaceX's previous IPO valuation record of 1.77 trillion US dollars and become the largest IPO in history. Wall Street is looking further into the future than usual to value this artificial intelligence company, with its valuation based on the revenue the company could generate two years later.

According to two people familiar with the company's financial situation, Anthropic expects its revenue to reach about 190 billion to 200 billion US dollars in 2028, a figure that has not been disclosed before.

Anthropic also plans to grant shares with additional voting rights to CEO Dario Amodei and other co-founders, a move that also prepares the company for its Wall Street listing. This equity structure is relatively common in the technology industry, aiming to ensure that founders have greater say in the company's development direction. Meta CEO Mark Zuckerberg and others also hold super-voting shares.

The pace of OpenAI's IPO is relatively unclear. On local time June 8, OpenAI CEO Sam Altman informed employees that OpenAI is expected to go public "within the next year". Altman said that the actual timing may be advanced or delayed, but "submitting the application now can give us greater flexibility". On the same day, OpenAI announced that it had secretly submitted the S-1 draft for its IPO to the U.S. Securities and Exchange Commission.

However, as OpenAI's losses continue to expand, many investors have doubts about the company's IPO valuation. Fidelity Securities wrote in a recent report that OpenAI's anchor valuation is "closer to 700 billion to 800 billion US dollars, rather than 1 trillion US dollars".

02

Both companies are competing to lay out in the enterprise AI track.

On May 4, Anthropic and OpenAI announced large-scale enterprise AI joint venture projects almost on the same day. OpenAI launched "The Deployment Company", partnering with TPG, Bain Capital, SoftBank and other institutions to build a 10-billion-US-dollar level enterprise AI deployment platform, providing customized AI implementation services for large groups. Anthropic, on the other hand, partnered with Blackstone and Goldman Sachs to set up a 1.5-billion-US-dollar enterprise AI service joint venture, focusing on the mass implementation of the Claude model for medium-sized enterprises to expand its own customer base.

Even after overtaking in revenue, Anthropic still faces multiple practical challenges. Its profitability is a phased result of a single quarter. Rising computing power costs, impact from open-source large models, and tightening budgets of enterprise customers may all erode profit margins. High dependence on enterprise customers also means that once downstream enterprises cut their AI budgets, revenue growth will quickly come under pressure. For OpenAI, the biggest challenge is how to balance growth, losses and capital market expectations.

Many industry investors believe that quarterly revenue only represents the present, and the real test is whether after listing, the company can continuously prove to the market that cutting-edge artificial intelligence can become a long-term and healthy business.

This article is from the WeChat official account "Jiemian News", written by Song Jiannan, and authorized for release by 36Kr.